Model reference fuzzy adaptive control of dissolved oxygen concentration

Biotechnological processes are notoriously difficult to control using conventional methods owing to their time-variant and non-linear characteristics. An alternative method is to utilise artificial intelligence such as fuzzy logic. A major difficulty with fuzzy logic control is rule elicitation: a lengthy process and specific to each application. To overcome this problem, the concept of self-learning and adaptive fuzzy logic controllers has been developed. In this paper, a different type of learning fuzzy control algorithm has been considered. The control method has been termed Model Reference Fuzzy Adaptive Control (MRFAC) indicating the method of adaptation. A reference model is used to provide performance feedback for automatically synthesising and modifying a fuzzy controller's rule-base on-line, as new information on how to control the system is gathered. A study is presented illustrating the performance of a MRFAC by demonstrating its ability to generate a rule-set based on predetermined criteria, when applied to a simulation of a Bakers' Yeast fed-batch fermentation. Results demonstrate how the learning mechanism controls the dissolved oxygen concentration throughout the duration of the fermentation process. Copyright (C) 2001 IFAC.